Intense competition
In recent days, the biggest news in the tech industry has been the two conferences held by OpenAI and Google.
Within 24 hours, the two AI giants, OpenAI and Google, made their moves one after another.
On May 14th, OpenAI CEO Sam Altman personally appeared at the event and, in a presentation lasting over 20 minutes, used the new multimodal model GPT-4o to awaken people's fantasies about the artificial intelligence system "Samantha" in the movie "Her".
The next day, Google announced an upgrade to Gemini, launching the "AI Toolbox" and integrating AI into its search engine, reshaping Android with AI. It is attempting to regain the initiative in the AI race. Over the past year, OpenAI has repeatedly outmaneuvered Google, releasing new technologies and products before Google.
Whether it's the technological competition between the two sides or OpenAI's partnership with Apple, it is evident that OpenAI, the current star in the AI field, poses an unprecedented threat to Google.
Of course, Google, with its search entry and the ace of the Android ecosystem, has a moat that is difficult to breach in the short term. However, in the AI era that is reshaping the industry landscape, facing the pursuit of the younger generation, tech giants like Google cannot afford to relax, or else it will mark the end of an era.
Battle
Despite being preempted by OpenAI, Google is clearly prepared.
While OpenAI introduced the articulate GPT-4o, Google also has its own Samantha—Project Astra.
"I've had this vision in my mind for a long time," said Demis Hassabis, head of Google DeepMind and head of Google's AI work. Over the years, he has been thinking and researching artificial intelligence, contemplating how to make machines more human-like.
In his vision, an AI agent must be able to understand and provide feedback on a complex and flexible world like a human. It should be able to see text and react to it, be proactive, educable, and have personality, so that it can converse naturally, without delay. Ultimately, all these imaginations became Project Astra. "It's that assistant," Hassabis said, "it's very useful, you're used to it being there when you need it."
From the demonstration video, it is clear that Project Astra still has a way to go to match Hassabis's vision. Even compared to GPT-4o, there is more latency and the communication is still not as natural. However, when Project Astra accurately said, "The glasses are on the desk, next to the red apple," it still elicited cheers from the audience.
The emergence of Project Astra is also seen as a positive response to GPT-4o.
This is not the first direct confrontation between OpenAI and Google. In February of this year, a week after Google's conference, OpenAI quietly dropped a bombshell, introducing its first text-to-video model—Sora. Although Sora has not been made available for public use to date, it still sent shockwaves through the entire tech industry On the bomb dropped by OpenAI, Google also counterattacked at the developer conference, releasing the AI media creation model Veo. According to reports, Veo can generate high-quality 1080p videos, relying on the latest text-to-image framework Imagen 3. These AI-generated videos can last for more than a minute, and Veo can also understand professional concepts in film production and visual technology, such as time-lapse shooting.
Furthermore, Google has integrated Gemini into almost all of its products, especially its core business - search, making its search capabilities even more powerful.
Despite the two conferences, one lasting only over 20 minutes and the other lasting 2 hours, for both companies, it was a technical showdown.
Anxiety
Behind the close combat between the two AI giants is the commercial anxiety of AI vendors.
In 2015, the relatively unknown OpenAI was established in Silicon Valley. At that time, it was still a non-profit organization. Relying on a $1 billion investment from Microsoft, it lived a modest life.
Since the launch of ChatGPT in November 2022, OpenAI made a splash, and its commercialization process has been accelerating. In April 2024, OpenAI's COO stated that the registered users of the enterprise version of ChatGPT have exceeded 600,000, compared to over 150,000 in January of this year.
Benefiting from the demand for AI technology from enterprises, it is reported that OpenAI's annual revenue in 2023 has exceeded $2 billion.
Under the AI wave triggered by ChatGPT, tech giants at home and abroad have all joined the game.
On February 6, 2023, Google announced the launch of the conversational AI chatbot Bard, the predecessor of Google's large model Gemini; Meta officially released Llama; in March of the same year, the American company Anthropic launched Claude; domestic companies quickly followed suit, with Baidu releasing Wenxin Yiyuan, Alibaba launching Tongyi Qianwen, Huawei's PanGu, and SenseTime's Ririxin.
Tech giants are flocking in, and for OpenAI to succeed in commercialization and seize the initiative, it must ensure its technological leadership and continuously introduce new and more powerful models. However, this path has become more challenging.
After more than a year of development, various vendors have successively launched large models comparable to GPT-4.
Even more challenging is that the pace of OpenAI's iterations has slowed down. It took OpenAI only over 4 months from the first generation ChatGPT to the release of ChatGPT. However, over a year has passed since the release of ChatGPT-4, and the much-anticipated ChatGPT-5 has yet to appear. While GPT-4o is indeed impressive, it is not enough to bring enough amazement to the entire industry Of course, the slowdown of OpenAI is mainly due to the objective law of large model iteration. The larger and more powerful the model, the higher the data and computing power requirements. Ultimately, all of these translate into cost pressures for OpenAI.
Compared to OpenAI's huge R&D costs, an annual revenue of over 2 billion USD is still a drop in the bucket. Sam Altman stated that due to the high costs of building and running its models, OpenAI is still operating at a loss. As OpenAI develops more complex models, expenses are expected to continue to exceed revenue growth. OpenAI may need to raise hundreds of billions of dollars to meet these costs.
In addition to external competition, OpenAI also faces "internal troubles." On May 15th, Ilya Sutskever, who has not appeared since the "palace struggle" incident at the end of last year, announced his departure, stating that he already has a new plan but cannot reveal details at the moment.
As the co-founder and chief scientist of OpenAI, Ilya Sutskever led the development of the GPT series of large models and the DALLE series of models. Although he did not attend the GPT-4o launch event, on the GPT-4o homepage, Ilya's name appears under the "additional leaders" section. The impact of Ilya Sutskever's departure on the commercialization of OpenAI remains difficult to assess.
Catching Up
As the technological gap gradually narrows, the competition among AI companies is also entering the second half.
At two recent events, OpenAI and Google tacitly began to focus on the edge. GPT-4o launched a Mac version application, while Google reshaped Android with Gemini.
China Galaxy Securities analysis believes that the release of GPT-4o heralds the beginning of a new round of competition in the AI Agent field, coupled with the integration of macOS desktop version ChatGPT, the edge AI revolution is imminent. In addition, the B-end enterprise service market may usher in new growth opportunities due to the ease of use and inclusiveness of GPT-4o.
In this new business opportunity, whoever can seize the initiative and occupy more terminals will be able to take the lead in the market.
In this regard, Google has a natural advantage. Google owns a series of applications such as Google Search, Gmail, YouTube, Chrome browser, etc., building a strong ecosystem.
In comparison, as the first company to introduce large models, although OpenAI has set a series of records under the AI dividend, there is still a considerable gap in user volume compared to Google.
According to Similarweb statistics, ChatGPT had a global visit volume of 1.8 billion times in April this year, an increase from March. Meanwhile, although Google's visit volume in April decreased compared to March, it still reached 83.49 billion times.
Not to mention, Google also has Android. At this year's developer conference, Google upgraded Android Studio, integrating more AI functions to facilitate developers to call the Gemini API for software development Jim Fan, the head of NVIDIA's AI studio, said that Google did one thing right, they finally made an effort to integrate artificial intelligence into the search box. In his view, distribution is Google's moat. In this context, Gemini doesn't necessarily have to be the best model in the world to become the most widely used model in the world.
In the war between OpenAI and Google, Google has a series of products like Android, while OpenAI lacks an Apple.
As the leader in smartphone shipments in 2023, iPhone shipments reached 234.6 million units last year, with a market share of 20.1%. If OpenAI's large models under its belt can truly replace "Siri" and become an AI smart assistant deployed on hundreds of millions of Apple devices, overturning the existing ecosystem in search experience and efficiency improvement, everything may become different for OpenAI